Comments/Ratings for a Single Item

I definitely don't wish to step on the toes of an existing game.

How about the name Swarm or Colony?

I am leaning toward Swarm, because it reflects the nature of the play.

After a quick websearch, I found that there are a lot of games based upon insects and that they have practically taken every possible name associated with them. Some even have duplicates.

But I think I might have one. 'Hivequeen'.

I'm also changing the name of the Warrior to Soldier. This will allow for better notation of the piece, and is probably a better term.

Names are a tricky thing.  I know the case of Dipole Chess, I asked Mark Steere if he minded me naming it.  Glad you didn't go with Swarm, as TWO games have that name.

A good place to check for names of games is on boardgamegeek.com .

It is definitely getting harder to come up with original names for games. Even using a really large thesaurus. ;-)

One of the goals in the development of this game was to create play which was extremely difficult for computers to quantify. The amount of depth-search necessary to arrive at a decent series of turns for this game would take days for most home units.

The field will become crowded in just a few opening turns. The middle game would consist of eliminating particular opposing pieces and denying the opponent 'breathing room'. The characteristics of the hexagonal field allows close placement of pieces which are out of direct reach of the attacking moves of the opponent.

And the value which the player places on particular pieces will influence the play. Some might prefer more Workers to assure the production phase of the turn, while others might concentrate on Soldiers to threaten the opponent's ranks. And repositioning a Soldier for an effective attack can take several turns, allowing the opponent time to make defensive positionings.

And though the Highborn has the slide move, it may not have opportunity to exercise it because of the crowded field. And the player may be reluctant to risk it.

And I know that someone might point out that if both players are determined, they could simply make moves which result in a simple back-and-forth of caputure and introduction. But this supposes that neither are trying to win the game. If this is the case, a maximum number of turns could be applied to the game with the calculation of the occupancy to determine the win. I would suggest 100 turns.

I have been considering an additional rule of allowing Workers to attack opposing Workers.

Each turn not only has the number of potential moves but also the number of potential introductions and their possible positions. This can be quite a large number.

Also the value judgement for each of these potentials can only be quantified by comparison to future positions. This comparison just in one turn can be rather larger. But such a comparison must be carried over into subseguent turns just to arrive at a possible series of moves and introductions which will give the player an advantage.

So just in four turns, the number of possibilities is quite large. First player turn X Second player turn X First player turn X Second player turn.  And this is still not sufficient for effective play. The depth-search would need to be at least eight plys just to offer a 'good' move.

It will be extremely difficult to implement this game for Zillions. And it would never play this game well. Because it would have to consider each of the introductions as a seperate move, whereas coding for the introductions as a single move would push its limitations.

But the human mind is able to calculate extremely fast with training. In fact, they can make intuitive leaps which a computer is not capable. The only advantage a computer has is accuracy. Most of the time. ;-)

| So just in four turns, the number of possibilities is quite large. 
| First player turn X Second player turn X First player turn X Second 
| player turn.  And this is still not sufficient for effective play. 
| The depth-search would need to be at least eight plys just to offer 
| a 'good' move.

Well, this is how it worked in the stone age, before they invented alpha-beta, null-move pruning and late-move reductions. I would be surprised if with these modern search techniques you would not reach 10 ply in a minute or so for this game. In normal Chess the top programs reach 20 ply in that time. That is in a game with on the average 40 moves per position, so according to your calculation, that should require them to search 40^20 or about 10^32 positions. Yet they can do it when searching about 1 million per second (so ~10^8 in total). They thus search about 4 times deeper than your equation suggests.

It is true that Zillions is rather inefficient in handling games with drops. It does not seem to reduce the search depth for them, as it should. But Zillions is of course hardly a standard to measure things by. Don't get me wrong, I think it is a miraculous product, and have great respect for its designers. But in being completely general, it is unavoidable that a lot of efficiency gets lost. A dedicated Chess program, even my 100-lines miniature micro-Max / Fairy-Max, totally crushes Zillions in the games that it is able to play (e.g. Capablanca Chess, Shatranj, Knightmate).

The reason I think HiveQueen would be easy to program is that most positions seem rather quiet to me: many pieces cannot capture at all. So it should be easy to evaluate by 'wood counting'.

Pruning techniques are commonly based on calculated values of the piece movements. But what if these values are 'outside' the movement?

For example: The Drone would be calculated at an extremely low value since its immediate movement does not result in the loss of material for the opponent. But it has another value which does not present itself until later in the game. Without a Drone, a player cannot promote a Highborn and thus risks losing the production phase of the turn.

Another thing that many chess programs have difficulty with is sacrifice. Without large look-aheads, they often do not detect the value of this maneuver. And in this game it may often require a sacrifice, or several sacrifices, in order to re-acquire the right of the production phase. For example, an opponent has reduced the player's Workers so that several pieces need to be loss to be able to introduce another.

So any pruning technique would need to continuely revisit previously discarded moves for evalution in relation to the examined ply.

Yes, you can create a program which would simply 'push' the pieces around. But would it ever be able to make tactical, or even strategic, sense of the play.

| Pruning techniques are commonly based on calculated values of the 
| piece movements. But what if these values are 'outside' the movement?

It seems your thinking is badly influenced by Zillions, which indeed seems to derive values of the pieces from their mobilities. It must, because it does not have any prior knowledge of the value of the pieces. When I write a dedicated program for a certain game, though, I measure the value of the involved pieces through asymmetric play-testing, and then program these values in 'by hand'. 

| For example: The Drone would be calculated at an extremely low value 
| since its immediate movement does not result in the loss of material 
| for the opponent. But it has another value which does not present 
| itself until later in the game. Without a Drone, a player cannot 
| promote a Highborn and thus risks losing the production phase of the 
| turn.

If having an extra Drone is as good as having 5 extra Soldiers, for whatever reason, the play-testing will reveal that. And I would simply award that score for the Drone being on the board, even if a Drone were a piece thas no moves at all.

These are not properties that make a game difficult for a computer to play. It might be difficult to acquire the knowledge that need to go in such a program, and it is unlikely that the program will be able to derive deep strategic motives in a few seconds with no other information than the rules. It takes thousands of high-quality Capablanca Chess games to derive a value for Archbishop and Chancellor. But it is not realistic to require that a program has to discover that knowledge for itself within a single game. If we teach people how to play Chess, we do tell them the piece values, even though they are not part of the rules at all.

Yes, pieces can be 'weighted' to allow for calculation. But this is an 'averaging' value. There would need to be constant update to this value to make it of significance.

And establishing a program which 'learns' would entail foreknowledge by the programmer of each and every potential which needs to be evaluated.

Just look to the programs of the game of Go. Very few are able to meet the ability of a master of this game. Even with structure and positional pattern recognition of the Go stones, the programs have difficulty in seeing the 'bigger picture'.

And the Mad Queen variant has crippled most game programmers in their analytical thinking. Seeing a game as only a series of moves and exchanges. (And yes, I know, that some programs look to positional patterns but they are few.)

Let me state that I do not believe that this particular game represents any serious hindrance to eventual quantification. Only that it represents a 'step' away from simple quantification.